HMGA1 acts as an epigenetic gatekeeper of ASCL2 and Wnt signaling during colon tumorigenesis

Abstract

Mutated tumor cells undergo changes in chromatin accessibility and gene expression, resulting in aberrant proliferation and differentiation, although how this occurs is unclear. HMGA1 chromatin regulators are abundant in stem cells and oncogenic in diverse tissues; however, their role in colon tumorigenesis is only beginning to emerge. Here, we uncover a previously unknown epigenetic program whereby HMGA1 amplifies Wnt signaling during colon tumorigenesis driven by inflammatory microbiota and/or Adenomatous polyposis coli (Apc) inactivation. Mechanistically, HMGA1 "opens" chromatin to upregulate the stem cell regulator, Ascl2, and downstream Wnt effectors, promoting stem and Paneth-like cell states while depleting differentiated enterocytes. Loss of just one Hmga1 allele within colon epithelium restrains tumorigenesis and Wnt signaling driven by mutant Apc and inflammatory microbiota. However, HMGA1 deficiency has minimal effects in colon epithelium under homeostatic conditions. In human colon cancer cells, HMGA1 directly induces ASCL2 by recruiting activating histone marks. Silencing HMGA1 disrupts oncogenic properties, whereas reexpression of ASCL2 partially rescues these phenotypes. Further, HMGA1 and ASCL2 are coexpressed and upregulated in human colorectal cancer. Together, our results establish HMGA1 as an epigenetic gatekeeper of Wnt signals and cell state under conditions of APC inactivation, illuminating HMGA1 as a potential therapeutic target in colon cancer.

Keywords: Colorectal cancer; Epigenetics; Oncology; Transcription.

Figure 1. Loss of a single Hmga1 allele mitigates colon tumorigenesis and prolongs survival in CDX2P-CreER^T2/Apc^fl/fl mice.

(A) Representative images (H&E) of proximal colon in CDX2P-CreER^T2/Apc^fl/fl mice with Hmga1 intact (Hmga1^+/+ top), heterozygous deletion (Hmga1^+/–, middle), or homozygous deletion (Hmga1^–/–, bottom) at survival endpoint necropsy (top: day 35 after TAM; middle: day 57; bottom: day 81). Scale bars: 250 μm. (B) Relative weight changes in CDX2P-CreER^T2/Apc^fl/fl models after TAM. (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, 1-way ANOVA). (C) Kaplan-Meier plot showing survival in CDX2P-CreER^T2/Apc^fl/fl mice with Hmga1^+/+, Hmga1^+/–, or Hmga1^+/+. (****P < 0.0001; Mantel-Cox test). (D) Relative colon weight to body weight in CDX2P-CreER^T2/Apc^fl/fl mice with Hmga1^+/+, Hmga1^+/–, or Hmga1^–/– (*P < 0.05; Hmga1^+/+ versus Hmga1^+/–, **P < 0.01, Hmga1^+/+ versus Hmga1^–/–; Tukey’s multiple comparison test following significance by 1-way ANOVA). (E) Proximal colon crypt depth in CDX2P-CreER^T2/Apc^fl/fl models (*P < 0.05, ***P < 0.01, ****P < 0.0001; 1-way ANOVA with Tukey’s multiple comparison test). Each shape (circle, square, triangle) corresponds to a different mouse (n = 2–3/genotype). The solid shapes show the mean from each mouse; the open, smaller shapes represent individual measurements/crypt (range = 9–13 crypts/mouse) at × 20 magnification.

Figure 2. Hmga1 haploinsufficiency decreases β-catenin and Ki67 in CDX2P-CreER^T2/Apc^fl/fl mice.

(A) Representative IHC images of nuclear HMGA1 (top), β-catenin (middle), and Ki67 (bottom) in CDX2P-CreER^T2/Apc^fl/fl models at 3 weeks after TAM. Scale bar: 200 μm. (B) Quantitative comparisons of IHC images (*P < 0.05, **P < 0.01; Tukey’s multiple comparison test following significance by 1-way ANOVA). Each shape (circle, square, triangle) corresponds to a different mouse [top bar graph (n = 2–3/genotype), middle bar graph (n = 3/genotype), bottom bar graph (n = 3/genotype)]. The solid shapes show the mean from each mouse; the open, smaller shapes represent individual values/field (range = 8–24 fields/mouse) at × 20 magnification.

Figure 3. Hmga1 deficiency decreases colon tumorigenesis.

(A) Representative IHC images of nuclear HMGA1 (top), β-catenin (middle), and Ki67 (bottom) in CDX2P-CreER^T2/Apc^fl/fl models at 3 weeks after TAM. Scale bars: 1,000 μm (top panel); 200 μm (lower panel). (B) Quantitative comparisons of IHC images (*P < 0.05, ***P < 0.001, ****P < 0.0001; Tukey’s multiple comparison test following significance by 1-way ANOVA). Each shape (circle, square, triangle, hexagon) corresponds to a different mouse (n = 3–4/genotype). The solid shapes show the mean values from each tumor; the open, smaller shapes represent individual values/field (range = 6–17 fields/tumor from 1–5 tumors/mouse) at × 20 magnification.

Figure 4. Hmga1 haploinsufficiency disrupts colon tumorigenesis induced by ETBF in APC^Min/+ mice.

(A) Body weights at necropsy after ETBF in Apc^Min/+ mice with intact Hmga1 or heterozygous Hmga1 (*P < 0.05, **P < 0.01, ***P < 0.001; student’s t test). (B) Representative images of methylene-blue stained colons to visualize tumors in Apc^Min/+/Hmga1^+/+ mouse (top) compared with Apc^Min/+/Hmga1^+/– mouse (bottom) at 11–12 weeks after ETBF. (C) Normalized tumor numbers in Apc^Min/+ models (*P < 0.05; Mann-Whitney test). (D) Kaplan-Meier plot showing survival in in Apc^Min/+ mice with intact Hmga1 or heterozygous (*P < 0.05; Mantel-Cox test). (E) Representative images (H&E left; IHC right; Scale bars: 200 μm) for HMGA1 (second column), β-catenin (third column), and Ki67 (right) in distal colon of Apc^Min/+ models at 11–12 weeks after ETBF. (F) Comparison of crypt depths (**P < 0.01) and IHC for nuclear HMGA1 (**P < 0.01), nuclear β-catenin (****P < 0.0001) and Ki-67(P = 0.16, unpaired student’s t test for each comparison) in Apc^Min/+ models. For crypt depth (left), each shape (circle, square, triangle, hexagon) corresponds to a different mouse (n = 3–4/genotype). The solid shapes show the mean from each mouse; the open, smaller shapes represent individual measurements/crypt (range = 9–16 crypts/mouse). For the IHC comparisons, each shape (circle, square, triangle, hexagon) corresponds to a different mouse (n = 3–4/genotype), the solid shapes show the mean value from each mouse; the open, smaller shapes represent individual values/field (range=9-19 fields/mouse) at x20 magnification.

Figure 5. Hmga1 haploinsufficiency decreases β-catenin and tumorigenesis induced by ETBF in APC^Min/+ mice.

(A) Representative images (H&E) of distal colon tumors in Apc^Min/+ models at 11–12 weeks after ETBF. (B) Representative images (IHC) of distal tumors in Apc^Min/+ models for HMGA1 and β-catenin at 11–12 weeks after ETBF. Scale bars: 200 μm. (C) Quantitative IHC comparisons of distal tumors in Apc^Min/+ models for HMGA1 (P = 0.09) and β-catenin (****P < 0.0001; unpaired student’s t test for both). Each shape (circle, square, triangle, hexagon) corresponds to a different mouse (n = 3–4/genotype). The solid shapes show the mean from each mouse; the open, smaller shapes represent individual values/field (range = 2–4 tumor fields/mouse) at × 20 magnification.

Figure 6. Loss of Hmga1 allele within colon epithelium decreases colon tumorigenesis induced by ETBF in Apc^Min/+ mice.

(A) Representative images of methylene-blue–stained colons of Apc^Min/+ mice (top) compared with Apc^Min with tissue-specific heterozygous Hmga1 deletion (middle) and tissue-specific homozygous Hmga1 deletion (bottom) at 11–12 weeks after ETBF. (B) Relative tumor numbers (%) in Apc^Min mice with intact Hmga1, tissue-specific heterozygous Hmga1 deletion, or tissue-specific homozygous Hmga1 deletion from 3 separate experiments; tumor numbers in control were assigned a value of 100 (*P < 0.05; Mann-Whitney test for both comparisons). (C) Representative images (H&E) of distal colon of Apc^Min/+ with or without tissue-specific Hmga1 deficiency models. Scale bars: 100 μm. (D) Distal colon crypt depths in Apc^Min/+ mice with or without tissue-specific Hmga1 deficiency. (**P < 0.01, ****P < 0.0001; Tukey’s multiple comparisons test following significance by 1-way ANOVA). Each shape (circle, square, triangle, hexagon) corresponds to a different mouse (n = 3–4/genotype). The solid shapes show the mean value from each mouse; the open, smaller shapes represent individual measurements/crypt (range = 9–19 crypts/mouse) at × 20 magnification. (E) Representative images (H&E) of distal colon tumors of Apc^Min/+ with or without tissue-specific Hmga1 deficiency.

Figure 7. HMGA1 expands colon stem cells and Paneth-like cells while depleting more differentiated cells in Apc-deficient colon crypts.

(A) UMAPs from scRNA-seq of crypt cells from CDX2P-CreER^T2 Apc^fl/fl mice with Hmga1^+/+ or Hmga1^–/–; shown together (left) or separately to highlight differences (center and right). (B) UMAP from scRNA-seq by cluster. Three distinct islands capture epithelial cell types (red circle), T cells (blue circle), and other immune cells (yellow). Imputed cell identities are designated by separate colors. TA, transit amplifying cells; EC, enterocytes. (C) Epcam, Lgr5, and other Wnt genes (Sox9, Ctnnb1) are enriched in the epithelial island. Single cell transcripts from both genotypes are shown. (D) Relative proportion of cell types in crypt cells by genotype (bar graph, left; Table, right). (Association between cell and HMGA1 status was evaluated by χ² test for each cell type versus all others).

Figure 8. Hmga1 deficiency alters cell state, decreasing stem and Paneth-like cell populations while expanding more differentiated cell populations in Apc deficient crypt cells.

(A) Pseudotime trajectory analysis estimated from scRNA-seq of CDX2P-CreER^T2 Apc^fl/fl crypt cells from the epithelial island with Hmga1^+/+ or Hmga1^–/–. HMGA1 deficient cells are more prominent in later stages of pseudotime (indicated by black ovals) compared with time 0 cells. (B) Cell states defined by the top 200 most differentially expressed genes on the trajectories from pseudotime analysis were assigned 0–4 and indicated by color on a trajectory plot (left) or bar graph (right). Note the skewing to cell states 3 and 4 in HMGA1 deficient cells. (C) Stem cells and enterocytes (ECs) imputed from scRNA-seq are shown on the trajectories to highlight the major differences between CDX2P-CreER^T2 Apc^fl/fl cells with intact HMGA1 or HMGA1 deficiency. HMGA1 deficient cells have increased ECs (blue) with decreased stem cells (violet). Bar graphs show relative cell frequencies (right); the top graphs show only stem and ECs, the bottom includes all cells with grey depicting cells that are not stem cells nor ECs.

Figure 9. HMGA1 activates gene networks within the crypt epithelial island involved in IFN signaling, inflammation, DNA repair, proliferation, and Wnt signaling.

(A) GSEA analysis (left) of single cell transcripts from the epithelial island reveals that HMGA1 activates pathways involved in inflammation (IFN-α, IFN-γ), DNA repair, and proliferation (MYC) while repressing pathways active in differentiated ECs (fatty acid metabolism, protein secretion); FDR ≤ 0.25. Enrichment plots (right) show HMGA1 networks in more detail, including genes involved in inflammation (IFN-α), DNA repair, and Wnt signaling. Normalized enrichment score (NES) and normalized P values are indicated. (B) IFN-inducible genes that mediate inflammatory signals, including IFN-induced transmembrane 1, 2, 3, (Ifitm1, 2, 3) genes, IFN stimulated gene 15 (Isg15), Stat1, Stat2, and cytokines (Ccl5, Cxcl9, Cxcl10) are activated by HMGA1. Dot plots depict gene expression (–0.4 to +0.4) and the proportion of cells (25%–75%) expressing each transcript within the epithelial island.

Figure 10. HMGA1 enhances chromatin accessibility at gene loci involved in proliferation, DNA repair, and inflammation.

(A) HMGA1 increases chromatin accessibility in crypt cell nuclei globally in CDX2P-CreER^T2 Apc^fl/fl mice. (B) HMGA1 enhances chromatin accessibility in promoter regions ranging from 0 to –3 kb upstream of the transcription start sites shown by average peak lengths. (****P < 0.0001; student’s t test). (C) HMGA1 enhances chromatin accessibility in promoter regions ranging from 0 to –3 kb upstream of the transcription start sites shown by number of significantly expanded peaks. (P < 0.0001; χ²). (D) HMGA1 enhances chromatin accessibility in gene sets involved in proliferation, inflammation, and metabolism. (E) GSEA pathways identified by intersecting ATAC-seq and scRNA-seq pathways with associated P values.

Figure 11. HMGA1 amplifies Wnt genes in Apc-deficient colon crypts, and these HMGA1-Wnt pathways are activated in human colon cancer.

(A) Dot plot of Wnt effector and receptor gene expression in crypt cells of CDX2P-CreER^T2 Apc^fl/fl mice with Hmga1^+/+ versus Hmga1^–/–, demonstrating that HMGA1 activates all Wnt effectors and many Wnt receptor genes. (B) Hmga1 is positively and strongly correlated with Wnt genes, including Wnt effector genes (Ascl2, Axin2, Tcf4, Ctnnb1, Myc, Ephb2, Sox9) and Wnt receptor genes (Lgr5, Lrp5, Fzd7) (Spearman’s rank correlation test). (C) Heatmap showing HMGA1 and WNT genes. HMGA1 correlates positively with ASCL2 and MYC (log scale) in human colon cancer. (D) HMGA1 and WNT genes in nonmalignant colon epithelium (n = 41) and human colorectal adenocarcinoma (n = 286) from TCGA (student’s t test). ^†HMGA1 and SOC9 expression were previously reported in ref. . TPM, transcripts per million. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 12. HMGA1 induces ASCL2 by directly binding to the promoter and recruiting activating histone marks.

(A) Silencing HMGA1 represses ASCL2 and decreases proliferation in SW620 and SW480 cells. Control cells were transduced with empty lentiviral vector versus shRNA targeting HMGA1. (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; student’s t test). (B) Silencing HMGA1 decreases clonogenicity in SW620 and SW480 (**P < 0.01, ****P < 0.0001; student’s t test). (C) Predicted HMGA1 binding sites 1–5 and region 6–7 in the ASCL2 promoter region shown with activating histone marks from SW620 and SW480 (GSE106921). (D) ChIP assay results at sites 1–5 and region 6–7 in SW620 cells from one representative biological replicate for HMGA1 and activating histone marks (H3K4me3,and H3K27ac). (E) ChIP assay results at sites 1–5 and region 6–7 in SW620 cells from one representative biological replicate for H3 and the repressive histone (H3K27me3). (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; student’s t test following significance by ANOVA).